The other day I found a really cool post on adding GIFs to ggplots, so I thought I’d give it a shot. Using spotifyr and ggplot2, I set out to plot Thom Yorke dancing across a plot describing Thom Yorke dancing. For those who have no idea who Thom Yorke is, he’s the frontman of Radiohead and Atoms for Peace, and he dances like this:

Mr. Yorke’s dance moves are already well GIF-ified, but finding a transparent GIF to add to a plot proved difficult. After some Googling, I decided to take these images of an animated Thom Yorke doing the Lotus Flower dance. I made them transparent with Preview for Mac and combined them into a GIF with imagemagick. I then read the GIF into R with the magick library.

library(tidyverse)
library(magick)

mini_thom <- image_read('thom_dance_og.gif')
mini_thom

To get data on danceability, I used spotifyr to pull track audio features for all albums by Radiohead, Atoms for Peace, and Thom Yorke’s self-titled solo act. I then filtered out non-studio albums and kept only danceability and track/album information.

“Danceability describes how suitable a track is for dancing based on a combination of musical elements including tempo, rhythm stability, beat strength, and overall regularity. A value of 0.0 is least danceable and 1.0 is most danceable.”

library(spotifyr)
library(tidyverse)

tots_thom <- map_df(c('radiohead', 'atoms for peace', 'thom yorke'), get_artist_audio_features)

non_studio_albums <- c('OK Computer OKNOTOK 1997 2017', 'TKOL RMX 1234567', 'In Rainbows (Disk 2)',
                       'Com Lag: 2+2=5', 'I Might Be Wrong', 'The Eraser Rmxs')

tots_thom <- filter(tots_thom, !album_name %in% non_studio_albums) %>%
    select(track_name, album_name, danceability, album_release_year) %>%
    unique() %>%
    arrange(-danceability)

library(knitr)

## Warning: package 'knitr' was built under R version 3.5.2

library(kableExtra)
tots_thom %>%
    kable() %>%
    kable_styling() %>%
    scroll_box(height = '500px')

Next I made a ridgeplot to show the danceability distributions by album, ordered by release date. I then saved the plot as a png for the GIF background.

library(ggridges)
library(lubridate)

# make label for plot
album_names_label <- tots_thom %>%
    arrange(album_release_year) %>%
    mutate(label = str_glue('{album_name} ({album_release_year})')) %>%
    pull(label) %>%
    unique

p <- ggplot(tots_thom, aes(x = danceability, y = as.character(album_release_year))) +
    geom_density_ridges() +
    theme_ridges(center_axis_labels = TRUE, grid = FALSE, font_size = 6) +
    theme(plot.title = element_text(face = 'bold', size = 14, hjust = 1.25),
          plot.subtitle = element_text(size = 10, hjust = 1.1)) +
    ggtitle('Have we reached peak Thom Yorke danceability?', 'Song danceability by album - Radiohead, Thom Yorke, and Atoms for Peace') +
    labs(x = 'Song danceability', y = '') +
    scale_x_continuous(breaks = c(0,.25,.5,.75,1)) +
    scale_y_discrete(labels = album_names_label)

ggsave(p, filename = 'danceplot.png', width = 5, height = 3)
background <- image_read('danceplot.png')
background

Frame by frame, I then added each individual image of my animated Thom GIF to the plot with image_composite(). To get him to dance across the x-axis, I moved the offset to the right by 100 pixels each frame.

frames <- map(1:length(mini_thom), function(frame) {
    hjust <- 200+(100*frame) # <- this makes him move along the x axis
    image_composite(background, mini_thom[frame], offset = str_glue('+{hjust}+400'))
})

Bringing it all together, I used image_animate(), setting loop = 0 for infinite looping.

image_animate(image_join(frames), fps = 5, loop = 0)